US20050027579A1 - Method and system for conducting fleet operations - Google Patents

Abstract

A fleet engine, a crew engine, a passenger engine and an integration engine that communicate with a distributed computer network via two-way communication channels to monitor and repair disruptions to schedules particularly in the airline industry. When a disruption occurs, the method will produce a plurality of solutions that are structurally different for evaluation by the controller or operations manager. The method of generating solutions includes two phases. A first phase with artificially relaxed costs and a second phase with costs that reflect the policies and actual costs of the relevant activities. Upon creating structurally different solutions, the method evaluates the solutions and presents summary information about the solutions to the operations manager. If a violation of a rule occurs in a solution, an alert is generated to notify a user of the rule violation.

Description

BACKGROUND OF THE INVENTION
• 1. Field of the Invention
• The subject disclosure relates to methods and systems for scheduling and rescheduling resources used to accomplish activities, and more particularly to an improved method and system for scheduling and rescheduling fleet vehicles before, during and after a disruption in operation.
• 2. Background of the Related Art
• Businesses in many areas utilize resources in complicated schedules to accomplish activities. For example, airlines, railways, buses, production lines and hospitals all have various resources including vehicles, machinery, floor space and personnel that must be coordinated on a grand scale. However, the schedules are subject to change based upon circumstances beyond the businesses control, profit objectives, policy changes and otherwise. When such disruptions occur, operations managers are typically unable to quickly and efficiently reschedule continuing operations without aid. In view of the above, several systems have been developed to aid decision making by proposing options for reallocating resources to achieve completion of the necessary activities. Such aids have been widely used and well understood in areas as diverse as airlines, railways, supply chains and logistics. Some examples are illustrated in U.S. Pat. No. 6,314,361, European Patent App. No. 1,195,670 and PCT Patent App. No. WO 02/097570 which are incorporated herein by reference.
• There are problems associated with the systems and methods of the prior art. Many algorithms are well know that apply operations to produce every combination in the neighborhood and pick the cheapest solution. However, this brute force approach may take unduly long as the size of the neighborhood may require execution of a large number of operations and a small “optimality gap” is acceptable to expedite selecting a solution. The “optimality gap” is the difference between a low cost solution that may be found quickly and an optimal solution that may take tremendous effort to find. Typically, prior art systems are designed to find a solution for a very large scale problem resulting from a major disruption. Moreover, such systems and methodology often take unacceptably long intervals to develop solutions which remain suboptimal. Although a plurality of solutions may be offered to the operations manager, the plurality tend to be minor variations upon the same solution rather than significantly different alternatives. There is a need, therefore, for an improved system and method which approaches optimally solving disruptions with a focus on the typical day to day minor disruptions and, yet is scalable to assist in very larger scale disruptions. Such an improved system would offer a plurality of structurally different solutions.
• Additionally, operations may involve multiple coordinated resources. For example, in the airline industry, operations managers often have to reschedule aircraft fleets as well as significant rescheduling of airline crews and passengers. Heretofore, an optimization aid used for one resource has been unable to interact with other optimization aids for the related resources. As a result, significant resources and valuable time are consumed pursuing rescheduling that is acceptable for utilizing one resource but completely unacceptable when the total impact is considered.
• For example, U.S. Pat. No. 6,314,361 to Yu et al. shows an optimization server1 that processes a request from a user for optimal solutions to a specific flight schedule disruption. In response to the request, the optimization server1 initiates an aircraft optimization engine3. The aircraft optimization engine3 processes the request and generates a set of solutions to overcome the disruption. In turn, the aircraft optimization engine3 initializes a crew optimization engine5 to determine whether the set of flight solutions are efficiently supported by flight and service crews. Many of the solutions or options produced by the optimization engine3, although reasonably optimized in consideration of aircraft utilization, turn out to be wholly unacceptable options when viewed in light of the ramifications upon crew and passenger inconvenience. Thus, critical resources and time are utilized to produce and evaluate solutions which are unacceptable and must be discarded. Accordingly, what is also needed is an integrated operations framework which allows information to be exchanged among different resource optimization engines prior to generating solutions to yield an overall optimum solution without expending critical resources on solutions directed to a portion of the solution without considering the whole.
SUMMARY OF THE INVENTION
• The present invention is directed to a method for generating a plurality of structurally different solutions with an optimization engine using recovery operations. The method includes the steps of periodically loading schedule data into the optimizaton engine and receiving a disruption specification and request for rescheduling from an operations manager. Upon receipt of the request, the optimization engine generates a plurality of solutions by using a local search algorithm in at least two phases. The first phase has an artificially relaxed cost and the second phase has normal costs that reflect policies and dollar values associated with corresponding activities. The plurality of solutions are evaluated and the most promising of the plurality of solutions are presented to the operations manager.
• It is an object of the disclosure to produce a plurality of solutions to major and minor disruptions as quickly as possible with as little change as possible while minimizing airline policy violations.
• It is another object of the disclosure to investigate variable constraints in a multi-phase approach to arrive at structurally different solutions.
• It is another object of the disclosure to limit the solution to one optimization engine's area of operation and simply evaluate the effect upon the other areas of operation.
• Still another embodiment of the present invention is a method for decomposing a disruption within an airline to facilitate generating optimal or near optimal solutions on a day of operations. The method utilizes fleet resources, crew resources and passenger schedules that are accessible by an integration server in communication with fleet, crew and passenger optimization engines. The crew and passenger optimization engines preprocess schedule information to generate initial data, wherein the initial data includes penalty cost values based upon hard and soft constraints plus costs based upon activities. The integration server receives a disruption specification and forwards the disruption specification to the fleet optimization engine. The initial data and the disruption specification are provided to the fleet optimization engine and used to prevent creating undesirable solutions by the fleet optimization engine. In effect, the initial data provides information about the flexiblity in the existing schedule of various operational areas, which will be valid for possible disruption specifications. During disruption processing one solver can request other solvers to generate solutions satisfying certain additional constaints. For example, a passenger solver can request a fleet upgrade (e.g., a larger aircraft) on a particular flight from the fleet solver to accommodate disrupted passengers.
• It is an object of the present disclosure to yield an overall feasible solution with minimal potential for an optimality gap. It is another object of the disclosure to guarantee the feasibility and effectiveness of solutions for the entire operation by integrating a plurality of components in an environment. It is another object of the disclosure to generate good quality solutions based upon lowest total penalty cost value.
• It is yet another object of the present disclosure to encapsulate each optimization engine's specific data and rules within the respective optimization engine. The optimization engines can communicate by exchanging solutions for evaluation across other optimization engines to yield a solution that is acceptable from an overall business perspective. Each optimization engine evaluates the impact of a solution within the corresponding area of operations. To reduce the number of solutions that an optimization engine must consider, any given optimization engine is advantageously designed to consider the other optimization engine's constraints and objectives.
• It should be appreciated that the present disclosure can be implemented in numerous ways, including without limitation as a process, an apparatus, a system, a device, a method, or a computer readable medium for applications now known and later developed. These and other unique features of the system disclosed herein will become more readily apparent from the following description and the accompanying drawings.
BRIEF DESCRIPTION OF THE DRAWINGS
• So that those having ordinary skill in the art to which the disclosed system appertains will more readily understand how to make and use the same, reference may be had to the drawings wherein:
• FIG. 1 an overview of an environment in which an embodiment of the present invention may be used.
• FIG. 2 is a flowchart illustrating a problem solving cycle in accordance with the subject disclosure.
• FIG. 3 is a flowchart illustrating in detail a method for generating solutions in accordance with the subject disclosure.
DETAILED DESCRIPTION OF PREFERRED EMBODIMENTS
• The present invention overcomes many of the prior art problems associated with optimization engines. The advantages, and other features of the system and method disclosed herein, will become more readily apparent to those having ordinary skill in the art from the following detailed description of certain preferred embodiments taken in conjunction with the drawings which set forth representative embodiments of the present invention.
• Referring toFIG. 1, there is illustrated a schematic representation of anenvironment100 in which the system and method of the present invention may be implemented. Theexemplary environment100 relates to the airline industry and, for simplicity, the following description relates to one airline operating theenvironment100. Theenvironment100 includes afleet engine102, a crew engine104, apassenger engine106 and anintegration engine108 which communicate with a distributedcomputer network110 via two-way communication channels. Note that the two-way communication channels are representative of a number of different communication channels known in the art such as telephone lines, optical cables, radio frequency, satellite and other means of transmission now known and later developed. When a disruption occurs, the subject system and method will produce a plurality of solutions that are structurally different for evaluation by the controller or operations manager.
• Eachengine102,104,106 stores a set of parameters related to resource utilization with associated costs. The costs are actual monetary costs and user selectable penalty value costs that reflect the user's business policies and objectives. Eachengine102,104,106 also contains feasibilty and legality information related to utilization of the resources. Theintegration engine108 exchanges data betweenother engines102,104,106 to yield integrated solutions.
• It is envisioned that each of theengines102,104,106,108 may incorporate one server or multiple servers. Multiple servers can cooperate to facilitate greater performance and stability of the subject invention by distributing memory and processing as is well known. In another embodiment, theenvironment100 will also include Disruption Manager servers [not shown] that are specific to particular operational areas. For example, an operations manager will be connected to a Fleet Disruption Manager server that provides access to relevant information and resources such as fleet engine data, operational alerts from the fleet engine and the like. In a preferred embodiment, theenvironment100 includes a Disruption Manager server for eachengine102,104,106,108. Distributedcomputer network110 may include any number of network systems well known to those skilled in the art. For example, distributedcomputer network110 may be a combination of local area networks (LAN), wide area networks (WAN), intranets or the Internet. The distributedcomputer network110 not only allows the components of theenvironment100 to communicate but components may be added and upgraded as desired. For example, a hub recovery engine [not shown] may be added to theenvironment100 and utilized in an embodiment of the subject disclosure as would be appreciated by those of ordinary skill in the art. The design of the interface of the distributedcomputer network110 places minimal requirements on components for facilitating integration. For example, the components may only need send and receive messages in a format which can be utilized by the other components. In a preferred embodiment, the distributedcomputer network110 only requires that the components read and write Extended Mark-up Language (“XML”) messages from an asynchronous queue.
• Auser interface112 is connected to thenetwork110 for providing the operations manager with access to theengines102,104,106,108. When a disruption to the schedule occurs, the operations manager will provide the particulars of the disruption in a request for a solution to theengines102,104,106,108 via theinterface112. The particulars of the disruption are referred to as the “disruption specification”. In a preferred embodiment, the operations manager can select a planning horizon within the disruption specification. The planning horizon is the period of time into the future that the solutions must consider. Thus, feasibility and legality are also considered within the planning horizon time frame. The part of the schedule that lies beyond the planning horizon is not checked for feasibility or legality. In a preferred mode, the default planning horizon is set as midnight on the current day of operation. Therefore, all schedule activities that begin within the planning horizon are verified for feasibility and legality. Theuser interface112 is designed to work in a multi user environment. A user can log in to theenvironment100 and receive a certain access level. For example, view only access will allow to the user to see the current state of the schedule and operational alerts, but not to modify such data.
• Upon receipt of the request for solution with disruption specification, theengines102,104,106,108 begin the process of providing a rescheduling solution. Theengines102,104,106,108 acquire the most recent schedule data from amemory storage system114 and perform operations to generate a rescheduling solution as described in more detail hereinbelow. Thememory storage system114 is connected to the distributedcomputer network110 by a two-way communication channel. Preferably, the data within thememory storage system114 is maintained automatically. In order to allow a hot start of theengines102,104,106,108, the data within thememory storage system114 is periodically downloaded to theengines102,104,106,108. Thus, when a request is received, the data synchronization merely involves the changes since the last refresh and, thereby, the data acquisition time is minimized. In a preferred embodiment, the data within thememory storage system114 is downloaded to theengines102,104,106,108 every two minutes.
• Referring now toFIG. 2, a graphical representation of the problem solving cycle is shown. When no disruptions are present, the operations manager monitors operations atstep200. When a disruption occurs, data relating to the disruption is entered viauser interface112 and an alert is generated as denoted byaction box205. The cycle proceeds to define the scope of the disruption atstep210. Defining the scope of the disruption includes determining the time frame, severity and resources affected by the disruption to generate a disruption specification. In a preferred embodiment, theintegration engine108 receives the data relating to the disruption, accesses the schedule data in thememory storage system114 and creates the disruption specification.
• The overall feasibility, legality and quality issues are controlled using theintegration engine108. In one embodiment, theintegration engine108 includes a submodule for storing and processing rules. Rules are resource specific and preferably encapsulated within each engine,102,104,106,108. The rules may be hard rules that cannot be violated or soft rules that can be violated by assumption of a corresponding penalty. The rules in theintegration engine108 are all soft such that overall best solutions are acquired by adjusting soft costs and parameters of theother engines102.104,106. By reviewing hard and soft rules for violations in theengines102,104,106, the solutions that violate feasibility, legality and quality policies can be discarded prior to being provided to the operations manager for review. In another embodiment, theintegration engine108 filters, discards and rates the solutions prior to presentation to the operations manager. Similarly, theother engines102,104,106 may utilize submodules for storing data and rules specific to the associated resource. In a preferred embodiment, the operations manager can introduce changes to the rules parameters within the rules submodules at short notice. By allowing changes to the rules submodules, changes to crew agreements, timetables, company policies, planning processes and the like can always be properly reflected in the solutions. Rules for aircraft are typically determined by the aircraft manufacturers with little, if any, variance there between so it is envisioned that changes would be infrequent. Example of hard rules are runway lengths, aircraft operating range, fuel capacity, number of seats available and the like whereas soft rules are maintenance intervals, turnaround time, curfews and the like.
• Atstep215, theintegration engine108 provides the disruption specification to eachengine102,104,106. Eachengine102,104,106 also preprocess based upon theschedule data114. The initial information is provided to theintegration engine108 for access by the other engines. Typical initial information would be penalty value costs associated with actions related to recovering from the disruption. For another example, thefleet engine102 generates initial information related to available standby aircraft, cancellation penalties, and preferred latest departure time for an aircraft affected by the disruption. The crew engine104 generates initial information related to tight connection constraints for crews, cancellation penalty value cost and crew limitations such as latest departure times, also called crew drop-dead limit. Thepassenger engine106 generates initial information related to flight values, passengers per class, connection constraints for passengers and cancellation penalty value costs. Thepassenger engine106 may request a fleet upgrade to a larger aircraft on a particular leg to accommodate disrupted passengers. Preferably, the initial information does not include the details required to generate the feasibility, legality and cost (both real and penalty value) data. For example, detailed passenger information is required by thepassenger engine106 during the preprocessing however thefleet engine102 and crew engine104 do not require such detailed passenger information.
• Preferably, the major preprocessing in theenvironment100 is performed during off peak hours such as during the night with smaller preprocessing tasks being done hourly. The operations manager can also manually trigger preprocessing. In short, a goal of preprocessing is to perform so-called sensitivity analysis, i.e. to inhibit the inherent flexibility in the schedule. The results of the preprocessing are used for processing disruption requests. Additional preprocessing is not performed after receipt of a disruption request mainly due to time constraints. Such initial information of additional constraints provides for efficient use of resources in generating solutions.
• After the preprocessing, theintegration engine108 may choose one ormore engines102,104,106 to generate solutions in view of the initial information. Moreover, as the selected engine or engines generate solutions, some solutions can be immediately discarded in view of feasibility, legality and excessive penalty problems identified by the other engines during preprocessing. Hence, the small amount of time spent preprocessing is more than saved by quickly discarding unacceptable solutions in view of initial information generated by theengines102,104,106. In short, even though the solution may be acceptable when viewed in the limited scope of one area of resources, the subject system and method quickly discards some of these solutions if the solution is undesirable when the overall impact is considered.
• In another embodiment, theintegration engine108 evaluates the disruption specification and identifies only one engine to generate solutions. In this exemplary embodiment, theintegration engine108 provides the disruption specification to thefleet engine102. Initially thefleet engine102 attempts to create a solution or solutions to the disruption without impacting other resources or schedules. Thus, evaluation of the solutions by theother engines104,106 will only take into account such actions as flight cancellations. As a result, theother engines104,106 may yield complex rescheduling solutions for their respective resources to accommodate the solutions. If the solutions must impact the other resources or schedules, the solutions are evaluated by the corresponding engines for feasibility, legality and penalty value cost atstep220. In still another embodiment, theintegration engine108 provides the disruption specification to one or more of theengines102,104,106 for each to generate solutions in parallel with or without preprocessing as desired.
• Still referring to step215 ofFIG. 2, the preferred method produces a plurality of structurally different solutions. For simplicity, the following disclosure is with respect to a single engine, thefleet engine102, generating solutions but it will be appreciated by those of ordinary skill in the art that the principles may be advantageously used with theother engines104 and106. Thefleet engine102 uses resources, e.g., aircraft, to complete activities called “legs”. A schedule is an assignment of an aircraft to a line of work wherein a line of work is an ordered set of legs. Thefleet engine102 uses a variety of operations to create a set of possible schedules or neighborhood. Typical recovery operations are flight delay, flight cancellation, aircraft type change and aircraft registration change. Each recovery operation incurs a user-specified penalty value cost. An aircraft having a fictitious line of work without any constraints may also be utilized by thefleet engine102 when rescheduling to represent cancelled or unassigned flights.
• Referring now toFIG. 3, a method for generating structurally different solutions using a two-phase search is described in more detail. Atstep300, thefleet engine102 has the disruption specification and up-to-date schedule data from thememory storage system114. Certain costs, preferably penalty value costs, are reduced to flatten the cost space. Atstep305, thefleet engine102 performs operations to develop a relaxed cost solution to the disruption until no further improvements to the solution are found. In effect, thefleet engine102 traverses the flattened cost space to find a local minimum within the neighborhood.
• Atstep310, the fleet engine enters the second phase of the procedure. The normal costs are reset to return the cost space to normal but the present location within the neighborhood does not change. Atstep315, thefleet engine102 begins operations from the relaxed cost solution within the cost space to determine a local minimum. In a preferred embodiment, partial banning is also used atstep315 to generate a plurality of local minimums. The local minimums become solutions for later evaluation atstep220 ofFIG. 2.
• Atstep320, thefleet engine100 decides whether or not to continue looking for additional solutions. The criteria for continuing the optimization process can be varied. For example, a certain number of variations may need to be run, a certain number of different solutions may be required or a time limit can stop the process. If the criteria for stopping has occurred, the fleet engine proceeds to step325 and the generation of solutions stops. Fromstep325, thefleet engine102 returns to step220 ofFIG. 2.
• Atstep320, if the stop criteria has not occurred, thefleet engine102 returns to step300. On subsequent passes throughstep300, different costs are modified. This varies the cost space in different ways so that instep305, thefleet engine102 performs operations which lead to different areas of the cost space. Then, as the fleet engine proceeds throughsteps310 and315, a structurally different solution is likely because the relaxed cost solution starting point is likely a different area within the cost space after the first phase. To further illustrate by example, thefleet engine102 may consider swaps, insertions and/or recombinations with no penalty value cost associated with missed maintenance on a certain aircraft to yield one structural solution on a first pass throughsteps300 and305. On subsequent passes throughsteps300 and305, different aircraft will have a relaxed penalty cost value to likely arrive at a different modified cost structural solution starting point forstep315.
• In another embodiment, additional phases are incorporated with relaxed and/or tightened costs prior to the resetting of the costs to determine the local minimum. In another embodiment, the first phase involves randomly picking points within the cost space and, during phase two, finding a local minimum to arrive at structurally different solutions.
• In still another embodiment, the multi-phase search approach is utilized to have the solutions reflect different recovery characteristics. A recovery characteristic would be to prefer swaps over delays by increasing the penalty value cost associated with delays. This favoring would be accomplished by having sets of temporary costs to be utilized in the first phase wherein the relationship between the sets would be orthogonal, i.e. drive the local search in different directions within the cost space. In addition to the initially preprocessed information, theengines102,104,106 may exchange a partial solution which can be completed by another engine to generate a solution with a required property. For example, while evaluating a particular recovery solution fromfleet engine102, thepassenger engine106 may identify a significant improvement of the corresponding passenger recovery solutions if a certain flight is upgraded to a larger aircraft. In this case, thepassenger engine106 may respond by suggesting an alternative solution, which includes the suggested flight upgrade. Accordingly, thefleet engine102 collaborates with crew engine104 andpassenger engine106 to arrive to the overall best solutions such that feasible operations within the crew engine104 andpassenger engine106 can be favored.
• Thefleet engine102 may use one or more of several search algorithms and techniques. The search algorithms may be the same for each phase or different. In a preferred embodiment, a steepest-descent algorithm is utilized to find solutions in both phases. In another embodiment, a SIMPLEX algorithm is utilized. It is envisioned that any commercially available, general purpose network and integer program optimization algorithm may be advantageously used by thefleet engine102 as would be appreciated by one of ordinary skill in the pertinent art.
• An additional variation is to precalculate which swaps may be beneficial. For example, swaps with large time differences will incur such a large penalty value that cancellations will remedy the problem on that line of work for a lower cost. Thus, cancellation costs can be used to prune the set of possible swaps and, in turn, decrease the size of the neighborhood. A smaller neighborhood increases search efficiency and prevents unacceptable solutions from being found. Another pruning technique is to only consider swaps or insertions that touch a line of work that is already affected by the disruption. In effect, the lines of work that do not need to be modified are left intact. It will be appreciated that the benefits of reducing the neighborhood by these and other techniques may be incorporated into the preprocessing by theengines102,104,106 in order to further increase the efficiency of the preferred method.
• Still another variation to arrive at different starting points within the cost space is to use purely cost variation schemes. By assigning different penalty value costs and/or expenses to certain activities and operations, different points in the cost space may be found. Still another variation includes substitution of equal and/or substantially equal alternatives. For example, several swaps may have the same cost. To generate different solutions, the order in which the swaps are tried may be randomized. This scheme works well for smaller neighborhood scenarios.
• Partial banning or taboo searching may also be utilized to arrive at structurally different solutions. Partial banning builds a tree of nodes wherein each node corresponds to a list of banned operations. The first or root node is created by generating a solution. For each operation used to generate the solution, a child node is attached. However, in each child node, one or more of the operations are banned. Thus, the options at each node will vary and, in turn, unique options will be generated. Partial banning may be integrated with or superimposed upon the multi-phase approach to further insure that structurally different solutions are generated. For example, within a particular iteration of a multi-phase approach, in which cancellations are favored over all other operations, partial banning will provide solutions with different flights cancelled. Hence, a multi-phase approach is used to generate structurally different solutions and partial banning provides still further variations. It is envisioned that the search methods and techniques above are not mutually exclusive.
• Multiple processors can be used in parallel to apply different methods to the same disruption. Similarly, different processors can run in parallel and apply the same technique but utilize different sets of costs, different partial banning or different favored operations to further reduce computation time and increase stability.
• Referring again toFIG. 2, atstep220, theintegration engine108 sends an evaluate solutions request to the engines that are working and receives the proposed solutions in response. In a preferred embodiment, the proposed solutions are presented to the operations manager for evaluation via theuser interface112. In another embodiment, the engines provide multiple solutions, ranked in order of ascending total cost. Preferably, the operations manager can specify an upper threshold on the total number of solutions to be proposed. To further expedite the optimization process, the operations manager can also specify an upper threshold on the number of solutions to be generated. Thus, whenever thefleet engine102 has found the requested number of solutions, thefleet engine102 compares the set of solutions in terms of total cost and proposes only the lowest cost solutions.
• The operations manager needs to be able to compare one solution to another. Revenue, operational cost, profit and performance are key figures of merit in comparing solutions, but these figures are not all represented in monetary terms so a user selectable weighting is applied to allow quantitative comparison of proposed solutions. The basic principle in the weighting selection process is that the operations manager specifies penalty cost values associated with different recovery operations. In order to determine the total cost of a proposed solution, the system and method sums the number and nature of such occurrences appearing in the solution to determine total penalty value cost. The actual costs are also determined. Actual costs include, without limitation, fuel costs, passenger loads and passenger values. A solution with a lower sum total of penalty cost value and actual costs will be deemed better than a solution with a higher sum. The total cost of a solution is user variable because of the selected penalty value costs assigned to certain events like delays.
• Preferably, the operations manager selects the penalty value costs based upon the policies and objectives of the airline. Accordingly, the objective of thefleet engine102 to find a relatively low cost or even optimal solution would reflect those policies and objectives. Preferably, thefleet engine102 provides a figure of merit for each solution to the operations manager wherein the figure of merit is the differential cost between the schedule before the disruption as compared to the schedule with the proposed solution implemented.
• Preferably, thefleet engine102 ranks the solutions that have been found and presents summary information in the form of an HTML document. The summary information includes the number of solutions found, the number of solutions displayed and the reason why thefleet engine102 stopped searching such as the time limit was reached or the specified number of options was found. If thefleet engine102 fails to propose a feasible solution,fleet engine102 may propose a partial solution. A partial solution would involve one or more infeasibilities and the solver will identify such in the summary information. Preferably, the user interface also can communicate with thefleet engine102, update databases such as theschedule data114, review alerts generated by the engines and the like. In a preferred embodiment, theengines102,104,106,108 automatically generate alerts.
• In another embodiment, theintegration engine108 provides the solutions to the crew engine104 andpassenger engine106 to determine feasibility and legality under the constraints therein as well as additional actual cost and penalty value cost information. Thus, theintegration engine108 collects information for the set of solutions based upon total overall cost and provides ranked overall solutions to the operations manager viauser interface112. The operations manager may also request eachengine102,104,106 to fully evaluation the schedule of one or more solutions for any rule violations, i.e. send an alert generation request. Theengine102,104,106 will evaluate the selected solutions for violations and generate an alert for each violation of the rules. The operations manager can further utilize the alerts to select and implement a solution.
• In another alternative embodiment, theintegration engine108 skips to step230 and automatically implements the overall lowest cost solution by updating theschedule data114 and utilizing the network to undertake remaining tasks such as notifying maintenance personnel, crews and passengers of schedule changes. It is envisioned that the notification may be made by updating Internet Web pages, automated telephonic communications by text messaging and otherwise, displays at the airport, electronic mail and other means of communication now known and later developed.
• Atstep225, the operations manager would take some or all of the solutions from thefleet engine102 and provide them to the crew engine104 andpassenger engine106. The crew engine104 andpassenger engine106 process these solutions to provide feasibility and legality verification plus additional cost information. Thus, the operations manager can select an option based upon the overall cost. If the operations manager deems a solution acceptable for implementation, the process proceeds to step230. If the operations manager does not find any of the solutions acceptable, the process can proceed to step215 for generating additional solutions by utilization of another method or further application of the same methods as described above. At step230, if a solution is acceptable, the solution is implemented. The airline reenters a monitoring operations mode atstep200 until another alert comes and the cycle continues.
• In another embodiment, the system and methods shown herein are useful as a simulation tool. The operations manager may modify the rules and/or penalty value costs to reflect different policies and input hypothetical disruptions. Review of the resulting solutions would allow quantitative assessment of the overall cost of certain policies during disruptions. Based upon these assessments, effective policies can be identified and implemented. For example, the operations manager can investigate different trade-offs such as between a quick recovery and a low operational cost, or between minimum changes and a stable operation.
• In another embodiment, the operations manager uses the system and methods herein to prepare cancellation plans based upon potential irregular operations. The cancellations plans can be evaluated by all resources in batch modes and real-time response would not be required. Thus, the optimality gap can be minimized or completely overcome. It is also envisioned that previous solutions may be stored for later use. If, at a later time, additional disruptions or the researched potential irregular operations occur, the stored solutions provide contingency plans for evaluation under the present circumstances without requiring additional generation of solutions by an optimization engine.
• While the invention has been described with respect to preferred embodiments, those skilled in the art will readily appreciate that various changes and/or modifications can be made to the invention without departing from the spirit or scope of the invention as defined by the appended claims.

Claims (33)

1. A method for generating a plurality of structurally different solutions with an optimization engine using recovery operations, the method comprising the steps of:

periodically loading schedule data into the optimization engine;

receiving a disruption specification and request for rescheduling from a user; and

generating a plurality of solutions by using a local search algorithm in at least two phases,

wherein a first phase has an artificially modified cost and a second phase has normal costs that reflect policies and monetary values associated with corresponding activities.

2. A method according toclaim 1, wherein the artificially modified cost is a penalty value cost associated with violation of a policy.

3. A method according toclaim 1, wherein the artificially modified cost is a monetary amount associated with an activity.

4. A method according toclaim 1, wherein the optimization engine operates on aircraft schedules and the artificially modified cost is set to favor a desirable operation over a less desirable operation as determined by one of a passenger and crew optimization engine such that a preferred operation by the passenger and crew optimization engines is more likely to be utilized by the portion of the plurality of solutions.

5. A method according toclaim 4, wherein the operations used by the aircraft optimization engine are flight delay, flight cancellation, aircraft type change and aircraft registration change.

6. A method according toclaim 1, further comprising the steps of:

receiving an alert generation request;

evaluating the plurality of solutions based upon at least one rule for violations of the at least one rule; and

generating an alert for each violation of the at least one rule.

7. A method according toclaim 1, wherein the second phase uses a different local search algorithm from the first phase.

8. A method according toclaim 1, wherein during a third phase, partial banning of operations is utilized.

9. A method according toclaim 1, wherein the local search algorithm starts from a randomly selected location.

10. A method according toclaim 1, wherein the local search algorithm is a steepest-descent algorithm.

11. A method according toclaim 1, further comprising the steps of:

monitoring operation of the schedule;

defining a scope of the disruption; and

pruning a neighborhood of potential solutions.

12. A method according toclaim 1, further comprising the step of selecting and implementing one of the plurality of solutions.

13. A method according toclaim 1, further comprising the step of synchronizing the schedule data upon receipt of the disruption specification.

14. A method according toclaim 1, wherein the schedule data is downloaded every two minutes.

15. A method according toclaim 1, further comprising the step of stopping processing after a predetermined amount of time.

16. A method according toclaim 15, further comprising the step of stopping processing after a user selected number of solutions are generated.

17. A method according toclaim 1, further comprising the steps of evaluating the plurality of solutions and providing at least a portion of the plurality of solutions to the user.

18. A method according toclaim 1, wherein the disruption specification includes a planning horizon.

19. A method for decomposing a disruption within an airline to facilitate generating optimal solutions on a day of operations involving a schedule coordinating fleet resources, crew resources and passengers, the method comprising the steps of:

providing an integration server in communication with fleet, crew and passenger optimization engines;

preprocessing based upon the schedule the crew and passenger optimization engines to generate initial data, wherein the initial data includes penalty cost values based upon hard and soft constraints plus costs based upon operations to the schedule;

providing the initial data to the fleet optimization engine;

receiving a disruption specification at the integration server;

forwarding the disruption specification to the fleet optimization engine; and

creating a plurality of solutions with the fleet optimization engine based upon the disruption specification and the initial data.

20. A method according toclaim 19, wherein the initial data includes cost data based upon hard and soft constraints.

21. A method according toclaim 19, creating a first plurality of solutions based upon the disruption specification wherein only fleet resources are rescheduled.

22. A method according toclaim 21, further comprising the step of terminating the method if one of the first plurality of solutions is acceptable.

23. A method according toclaim 19, further comprising the step of providing the plurality of solutions generated by the fleet optimization engine to the crew and passenger optimization engines for cost evaluation.

24. A method according toclaim 19, wherein the initial data generated by the passenger optimization engine includes flight values, passengers per class, connection constraints for passengers and cancellation penalty value costs.

25. A method according toclaim 19, wherein the initial data generated by the crew optimization engine includes crew limitations.

26. A method according toclaim 25, wherein a crew limitation is latest departure times.

27. A method according toclaim 19, further comprising the step of preprocessing on the schedule by the fleet optimization engine to generate initial data that is provided to the crew and passenger optimization engines.

28. A method according toclaim 27, wherein initial data generated by the fleet optimization engine is provided to the crew and passenger.

29. A method according toclaim 28, wherein the initial data generated by the fleet optimization engine is latest departure times for aircraft directly affected by the disruption.

30. A method according toclaim 19, further comprising the steps of:

generating, by the fleet optimization engine, a plurality of solutions by using a local search algorithm in at least two phases, wherein a first phase has an artificially relaxed cost and the second phase has normal costs that reflect policies and dollar values associated with corresponding activities;

evaluating the plurality of solutions; and

providing at least a portion of the plurality of solutions to the integration server.

31. A method according toclaim 19, further comprising the steps of generating, by the fleet optimization engine, a plurality of solutions by using a local search algorithm from random points within a cost space based upon the schedules.

32. A system for generating a plurality of structurally different solutions with using recovery operations, the system comprising:

first means for loading schedule data into the optimization engine;

second means for receiving a disruption specification and request for rescheduling from a user; and

third means for generating a plurality of solutions by using a local search algorithm in a first and second phase, wherein the first phase has a modified cost and the second phase has costs that reflect policies associated with corresponding activities.

33. A system as recited inclaim 32, wherein the first means is a user interface in communication with memory, and the second and third means are an optimization engine in communication with the first means.

Images